Downloaded from https://www.cambridge.org/core.

Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

Political Science Research and Methods (2019), page 1 of 8

doi:10.1017/psrm.2019.18

RESEARCH NOTE

Policy over party: comparing the effects of candidate

ideology and party on affective polarization
Yphtach Lelkes*
Annenberg School for Communication, and Political Science (Secondary), University of Pennsylvania, 3620 Walnut Street.
Philadelphia, PA 19104, United States
*Corresponding author. Email: ylelkes@upenn.edu

(Received 30 July 2018; revised 7 December 2018; accepted 20 December 2018)

Abstract
At least two theories have been offered that explain the rise of affective polarization. Some scholars, relying
on social identity theory, argue that as the relevance of party identification increased, Americans became
more likely to see their in-party in positive terms and the out-party in negative terms. Other scholars
argue that affective polarization is a reaction to increasingly extreme political actors. This study seeks
to arbitrate between these two theories of affective polarization through a survey experiment which
asks respondents to rate candidates whose party (or lack thereof) and ideology (or lack thereof) is ran-
domly assigned. In line with the policy-oriented view of affective polarization, respondents reacted far
more strongly to ideology than party, especially if it was the ideology of the member of the out-party.

Keywords: Public opinion

Interparty animosity, or affective polarization, has increased in America over the past few decades
(Iyengar et al. 2012). Partisans today are more likely to say they dislike members of the other
party, avoid interacting with them, and even take material losses in order to avoid helping the
other side (for a review, see Iyengar et al. 2019). While there is a consensus that America
today is affectively polarized, there is less consensus as to how it got to be that way.1
Scholars have offered at least two competing explanations for the rise in affective polarization.
Some argue that partisan identities have been the main drivers of affective polarization: As
Americans have sorted into different homogeneous tribes (Mason 2018; Mason and Wronski
2018), partisan identities have become more salient, and, as a consequence, intergroup animosity
has increased. Others have argued that ideology plays a major role. In particular, elite polarization
and partisan sorting have driven Americans to see large policy differences between their side and
the other side (Fiorina 2016; Rogowski and Sutherland 2016; Webster and Abramowitz 2017),
making affective polarization a logical outcome. The ideology and party identity perspectives
have largely developed in parallel: Some studies have examined whether partisan identity is
related to affective polarization (e.g., Mason 2018), while other studies have examined whether
policy preferences matter (e.g., Rogowski and Sutherland 2016; Webster and Abramowitz
2017). No study has compared the predictive power of each perspective.
The current study examines whether the ideology of a political candidate or his or her party
produces more affective polarization. Respondents were randomly assigned to learn about a can-
didate’s ideology (which varied in extremity), their partisanship, or both. Compared to a control
condition, learning that a candidate held ideologically extreme positions generated feelings that
1
It should be noted that Klar et al. (2018) raise another possibility: affective polarization is a measure of growing distaste
towards partisanship, rather than a measure of interparty animosity.
© The European Political Science Association 2019.
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

2 Yphtach Lelkes

were 2–3 times as large as only learning about their partisan identity. These effects were moder-
ated by the respondent’s ideology. Respondents who hold moderate attitudes are not affected by
the candidate’s positions, while ideologues punish out-party extremists and reward in-party
extremists.
In addition to comparing the main effects of ideology and party, this study also examines
whether these effects depend on whether a candidate is a member of the in-party or out-party.
Recent literature has argued that the current political environment is marked by “negative parti-
sanship” (Abramowitz and Webster 2018). That is, people care more about the other party than
their party; others find that people are motivated primarily by their in-group (Lelkes 2018;
McConnell et al. 2018). I find that the effects of ideology and party identity are indeed asymmet-
ric: information about the out-party yields effects that are between 50 and 100 percent bigger
than information about the in-party. However, in-party information still yields substantively
large effects on feeling thermometer scores.
While both the party and the ideology of candidates exacerbate affective polarization, the
impact of ideology, especially that of an out-group member, is far greater than some literature
implies. This also implies that while Americans themselves may be “ideologically innocent”
(Kinder and Kalmoe 2017), they still judge political elites by their political ideology.

1. What drives affective polarization?

Two lines of research seek to explain the rise of affective polarization. First, some argue that par-
tisanship as a social identity is the principal driver of affective polarization (Iyengar et al. 2012;
Mason and Wronski 2018). As political parties have become more homogeneous in terms of the
race, religiosity, ideological identity, and geography of their constituents, partisanship became an
increasingly important social identity (Mason and Wronski 2018). As partisans no longer hold
identities that cross across party lines, they become more strongly attached to their own group
and hostile to the other side. This line of work implies that merely knowing someone is from
the other side is enough to generate animus (Iyengar et al. 2012).2
A few scholars have offered an alternative to the social identity approach and claim that, at its
root, affective polarization is ideological. Two sets of theories link affective polarization to ideol-
ogy. Some scholars believe that affective polarization is causally linked to partisan-ideological
sorting (Bougher 2017; Webster and Abramowitz 2017). Over the past half-century or so, par-
tisan identities have become increasingly aligned with ideological position (Levendusky 2009).
That is, Republicans are more likely to hold conservative positions and Democrats are more
likely to hold liberal positions than they were in the past. Bougher (2017) and Webster and
Abramowitz (2017) argue that this process has increased affective polarization, but Lelkes
(2018) shows that the two processes are only weakly linked: both the unsorted and sorted
have become more affectively polarized; lagged affective polarization predicts sorting at all
knowledge levels, while lagged sorting only predicts affective polarization among the most
politically knowledgeable.
Others argue that rising elite ideological extremity has driven up partisan animosity (Rogowski
and Sutherland 2016; Webster and Abramowitz 2017) as it raises the stakes of politics. Having a
political extremist in office means that change might not be incremental, but rather dramatic.
While both Rogowski and Sutherland (2016) and Webster and Abramowitz (2017) are important
contributions to the literature, neither let us compare the relative effects of the party identity of a
politician versus his or her political positions. Rogowski and Sutherland (2016) experiment does
not identify the candidate’s party and identifies ideology with their placement on an 11-point
ideological self-identification scale, which potentially conflates identity with substantive ideology.
2
It should be noted that there are multiple variants of the social identity explanation to affective polarization (Johnston
2018).
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

Political Science Research and Method 3

Webster and Abramowitz (2017) study gives respondents the operational ideology of the polit-
ician but does not vary the politician’s party identity, which makes a comparison of the effect
of party versus ideology impossible.

2. Data and methods

Partisans were recruited from the Qualtrics online panel; pure independents were not included
in the sample (see Appendix for demographics). Although the panel is not representative of
the US population, experimental studies find equivalent results with representative and non-
representative samples (Berinsky et al. 2012; Mullinix et al. 2016) Two experiments were run,
one in which respondents were asked to rate someone from their in-party (N = 1526) and one
in which respondents were asked to rate someone from the other party (N = 4272). For the
sake of presentation, data from the two experiments were concatenated. Roughly 400 respondents
took part in the second experiment (out-party) as well as the first experiment (in-party).
Out-party data from these respondents were excluded from the analysis, although the results
are substantively the same if they are left in the sample.
Within each of these experiments, respondents were randomly assigned to read about one of
six fictitious candidates who was ostensibly a member of the respondent’s in-party or out-party
(see Appendix for wording). In the control condition, they read a short biography, and were then
asked to rate the candidate on a feeling thermometer scale. In the partisan identity condition, they
were also told whether the candidate was a Republican or a Democrat. In the moderate-no party
id condition, they were told that the candidate held a host of moderate issue positions but not the
candidate’s party. If the candidate was a Republican, they took the issue positions of Susan Collins
in 2017, the most moderate Republican in the Senate (DW-Nominate First Dimension Score of
0.11 as of writing). If the candidate was a Democrat, they took the issue positions of Joe Manchin,
the most moderate Democrat in the Senate in 2017 (DW-Nominate First Dimension Score of
−0.06 as of writing). In the extreme-no party id condition, they were told that the candidate
held a host of extreme issue positions but not the candidate’s party. If the candidate was a
Democrat, they took the issue positions of Elizabeth Warren, the most liberal Senator in 2017
(DW-Nominate First Dimension Score of −0.78 as of writing). If the candidate was a
Republican, they took the issue positions of Mike Lee, the most conservative Senator
(DW-Nominate First Dimension Score of .92 as of writing). In the moderate-party id and
extreme-party id conditions, they were also told whether the candidate was a Democrat or a
Republican.
For instance, if the respondent was assigned to the extreme-party id, they would see the
following:

William Richardson is a Republican. He is a 53-year-old businessman who has been married

for 34 years and has two sons. He earned his college degree in 1974 and then served for ten
years in the state legislature. After that, he was a co-founder of the state’s public affairs tele-
vision station. In addition, he is a member of the board for the state history museum.
Richardson strongly opposes government-run universal health care. He believes that the
federal government should not be able to mandate health care and favors a free-market sys-
tem. In addition, this candidate is strongly pro-life and wants to prohibit federal funding for
abortions. Lastly, he supports legal immigration only and wants tougher enforcement of
immigration laws. He wants the federal government to better secure the southern border
and not grant amnesty for those who enter the US illegally.
On a scale ranging from 0 to 100, where a score of 0 means you feel “cold” towards this
candidate, a score of 50 means you do not feel “particularly warm or cold” towards this can-
didate, and a score of 100 means you have “warm” feelings towards this candidate, how
would score your feelings towards this candidate?
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

4 Yphtach Lelkes

To estimate the effect of each condition, feeling thermometer scores towards the candidate were
regressed on treatment condition. To determine whether ideological divergence predicted
changes in affective polarization, feeling thermometer scores were regressed on a series of
treatment dummies, ideological extremity, and the interaction between the two.

3. Results
The effect of learning that a candidate had extreme political preferences was much larger than the
effect of learning about the candidate’s political party. Compared to those in the control group,
learning a candidate from your side holds extreme policy preferences increases feeling thermom-
eter scores roughly 3.5 times as much as learning that the candidate is from the same party
(b = 14.59, se = 1.92, p < 0.001 versus b = 4.35, se = 1.95, p < 0.05, respectively; Rows 4 and 1 in
Figure 1). Similarly, being told that a candidate holds extreme policy preferences associated
with the other party decreases feeling thermometer scores roughly 2.5 times as much as learning
that the candidate is from the other party (b = −24.99, se = 1.25, p < 0.001 and b = −10.15, se =
1.25, p < 0.001, respectively; Rows 6 and 2 in Figure 1).
The absolute values of these coefficients also indicate that the effect of receiving an out-party
cue is far larger than the effect of receiving an in-party cue. The effect of learning that a candidate
from one’s side held extreme attitudes was roughly 1.7 times smaller than learning that a candi-
date from the other side held extreme attitudes. The effect of learning that a candidate identified
with the same party the respondent was roughly 2.5 times smaller than learning that a candidate
was from the other side.
Additionally, learning that a candidate held positions associated with a moderate from one’s
own side slightly decreased feeling thermometer scores (b = −7.64, se = 1.90; Row 3 in
Figure 1), while learning that the candidate policy positions associated with a moderate from
the other side had no effect on feeling thermometer scores (b = 1.50, se = 1.24, p = 0.22; Row 5
in Figure 1). The combination of learning about party and policy were the same as learning
about policy alone (Rows 7–10 in Figure 1).
The impact of policy cues was particularly strong among respondents who held strong policy
positions. If a respondent was told that candidate held extreme in-party views, ideologically extreme
respondents registered feeling thermometer ratings 35.89 (se = 11.29) points warmer than ideologic-
ally neutral respondents (Row 2, Column 2 in Figure 2). If the respondent read about an ideologic-
ally extreme out-party candidate, ideologically extreme respondents rated the candidate 25 points
lower than ideologically neutral respondents when only the ideology was given (−25.25, se = 3.21,
p < 0.01; Row 3, Column 2 in Figure 2) or 22 points lower when both the ideology and cue were
given (b = −21.76, se = 3.22, p < 0.01; Row 5, Column 2 in Figure 2).
The respondent’s ideological extremity did not moderate the impact of receiving an inparty-
cue (Row 1, Column 1 in Figure 2), an out-party cue (Row 1, Column 2 in Figure 2), or learning
that the candidate held in-party moderate (Row 2, Column 1 in Figure 2) or out-party moderate
(Row 3, Column 1 in Figure 2) views. Ideologically extreme respondents were more likely to rate
candidates negatively if they were told that the candidate identified with the respondent’s party
and held moderate viewpoints (b = −24.32, se = 10.76, p < 0.05 (Row 4, Column 1 in Figure 2).

4. Discussion
These results support theories of affective polarization that privilege ideological preferences over
social identity (Bafumi and Shapiro 2009; Rogowski and Sutherland 2016; Fiorina 2017; Webster
and Abramowitz 2017), while raising questions about studies that argue that party dominates pol-
icy (e.g., Converse 1964; Cohen 2003; Kinder and Kalmoe 2017; Mason and Wronski 2018). For
instance, Mason argues that Americans dislike each other without necessarily disagreeing with
one another. What matters, in her framework, is a shared social identity. The results of the
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

Political Science Research and Method 5

Figure 1. The Impact of Candidate Party and Ideology on Feeling Thermometer Scores.

current study suggest that a shared social identity had a relatively small impact on a person’s feel-
ings towards a candidate, while disagreements over policy had substantively large effects. Policy
preferences matter, especially among respondents who held extreme policy preferences: Extreme
respondents react most favorably to extreme candidates of their own side and most negatively to
extreme candidates from the other side. Respondents with the least extreme attitudes were not
moved by the policy preferences of a candidate.
Relatedly, this work points to the importance of ideology in structuring political evaluations.
Many have assumed that party labels are enough to overcome “informational shortfalls about pol-
itics” (Sniderman 2000, p. 81). As the present research shows, while party labels indeed help
voters distinguish between candidates, party labels alone do not provide enough information
to overcome not knowing policy positions. In fact, knowing both issue position and party labels
offers no additional information beyond knowing issue positions alone. These results also imply
that even in today’s polarized environment, voters are not simply projecting the most extreme
policy positions onto candidates.
I also find that people react more strongly to the ideology and partisanship of out-party can-
didates than in-party candidates, in line with theories of negative partisanship (Abramowitz and
Webster 2018). These findings do not indicate that in-party cues and ideology do not matter, only
that the candidate’s ideology and out-group behavior matter more.
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

6 Yphtach Lelkes

Figure 2. The Impact of Candidate Party and Ideology on Feeling Thermometer Scores by Respondent Ideological
Extremity.

This work also points to a tension in the American politics literature: Americans are not ideo-
logues (Kinder and Kalmoe 2017), yet, as demonstrated by this research, they also care deeply
about the ideological preferences of political candidates. How do we reconcile these seemingly
antithetical empirical findings? One possibility is that although most people are not ideologues,
they are increasingly aware of the ideological reputations of the political parties (Sniderman and
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

Political Science Research and Method 7

Stiglitz 2012). This suggests that citizens are aware of the policy positions that would make a can-
didate a “good partisan,” even if they do not hold those positions themselves.
In addition to contributing to the role of policy and party in explaining affective polarization,
this study also contributes to the literature on information and candidate evaluation. Alvarez
(1998, p. 109) argues that “the more uncertain a voter is about the candidate, the less likely
they are to support the candidate.” Respondents in the control condition, who received no sub-
stantive information were more positive about the candidate than respondents in all other con-
ditions except for those in the in-party cue, extreme in-party policy preferences or the mix of the
two. Counter Alvarez (1998), respondents did not “prefer the devil they know more about to the
devil they know less about.”
While this study helps us understand the roots of affective polarization, there are important
caveats. This work says nothing about the validity of the social sorting process and its effects
on affective polarization (Mason and Wronski 2018). It may be that social sorting and the result-
ing identities are the distal reason that ideology exerts such strong effects. Strong identifiers may
believe that strong co-partisan ideologues signal strong party allegiance, while strong out-partisan
ideologues are a particular threat.
While the use of feeling thermometers as a measure of affective polarization has been validated
(Druckman and Levendusky 2018), they are ultimately summary evaluations, and can be used, for
instance, as an indicator of candidate utilities (e.g., Palfrey and Poole 1987). Hence, there is some
ambiguity in whether the use of a feeling thermometer in this experiment is indicative of affective
attachment or something else. Future studies could implement other measures of affective polar-
ization, including behavioral outcomes (McConnell et al. 2018).
This study also does not shed light on the role of a third possible pathway to affective polar-
ization: the information environment. Partisans, inadvertently and advertently, are now more
likely to be exposed to vitriolic partisan messages than in prior years, and people with stronger
identities, perhaps formed by the social sorting process described by Mason and Wronski (2018),
are more receptive to this information. Exposure to this information has also been linked to
affective polarization (e.g. Levendusky and Malhotra 2016).
Although this design is similar to that of many studies that examine political attitudes (Tomz
and Van Houweling 2009; Sniderman and Stiglitz 2012; Rogowski and Sutherland 2016; Webster
and Abramowitz 2017) in that it uses a fictional candidate, a cautious view of the external validity
of this study should still be taken. Note that Rogowski and Sutherland (2016) find similar results
in their experimental study with fictional candidates as well as their observational studies with
real candidates.
This study indicates that, if we are to decrease affective polarization in the US, we need
politicians that are politically moderate. Unfortunately, voters prefer politicians of their own
party that are politically extreme. This incentivizes extreme political candidates, which will
only exacerbate current tensions.
Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2019.18

Acknowledgments. I thank Nathan Kalmoe, Paul Kellstedt, Neil Stenhouse, and two anonymous reviewers for their com-
ments. I thank Dave Vanette for running the study on the Qualtrics panel, and Eric Selzer for valuable research assistance.

Author ORCIDs. Yphtach Lelkes, 0000-0003-1805-056X.

References
Abramowitz AI and Webster SW (2018) Negative partisanship: Why Americans dislike parties but behave like Rabid
Partisans. Political Psychology 39, 119–135, ISSN: 0162895X.
Alvarez RM (1998) Information and Elections. Ann Arbor: University of Michigan Press.
Bafumi J and Shapiro RY (2009) A new partisan voter. The Journal of Politics 71, 1–24.
Downloaded from https://www.cambridge.org/core. Kean University, on 25 Jul 2019 at 08:07:34, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/psrm.2019.18

8 Yphtach Lelkes

Berinsky AJ, Huber GA and Lenz GS (2012) Evaluating online labor markets for experimental research: Amazon.com’s
mechanical turk. Political Analysis 20, 351–368.
Bougher LD (2017) The correlates of discord: identity, issue alignment, and political hostility in polarized America. Political
Behavior 39, 731–762.
Cohen GL (2003) Party over policy: the dominating impact of group influence on political beliefs. Journal of Personality and
Social Psychology 85, 808.
Converse PE (1964) The nature of belief systems in mass politics. In D Apter (ed.), Ideology and Discontent. New York: The
Free Press of Glencoe, pp. 206–261.
Druckman J and Levendusky MS (2018) What do we measure when we measure affective partisanship? Public Opinion
Quarterly. Paper in press.
Fiorina M (2016) Has the American public polarized? Contemporary American Politics 2, 283–301.
Fiorina M (2017) Unstable Majorities: Polarization, Party Sorting, and Political Stalemate. Stanford, CA: Hoover Press.
Iyengar S, Sood G and Lelkes Y (2012) Affect, not ideology: a social identity perspective on polarization. Public Opinion
Quarterly 76, 405–431. ISSN: 0033-362X.
Iyengar S, Lelkes Y, Levendusky M, Malhotra N and Westwood SJ (2019) The origins and consequences of affective polar-
ization in the United States affective polarization: an outgrowth of Partisan social identity. Annual Review of Political
Science, 1–35. https://www.annualreviews.org/doi/abs/10.1146/annurev-polisci-051117-073034
Johnston CD (2018) Authoritarianism, affective polarization, and economic ideology. Political Psychology 39, 219–238. ISSN:
14679221.
Kinder DR and Kalmoe NP (2017) Neither Liberal nor Conservative: Ideological Innocence in the American Public. Chicago,
IL: University of Chicago Press.
Klar S, Krupnikov Y and Ryan JB (2018) Affective polarization or Partisan Disdain? Untangling a dislike for the opposing
party from a dislike of Partisanship. Public opinon quarterly.
Lelkes Y (2018) Affective polarization and ideological sorting: a reciprocal, Albeit weak, relationship. The Forum: A Journal of
Applied Research in Contemporary Politics 16, 67–79.
Levendusky MS (2009) The partisan sort: how liberals became Democrats and conservatives became Republicans.
Levendusky M and Malhotra N (2016) Does media coverage of partisan polarization affect political attitudes?. Political
Communication 33(2), 283–301.
Mason L (2018) Uncivil Agreement: How Politics Became our Identity. Chicago, IL: University of Chicago Press.
Mason L and Wronski J (2018) One tribe to bind them all: How our social group attachments strengthen Partisanship.
Political Psychology 39, 257–277. ISSN: 0162895X
McConnell C, Margalit Y, Malhotra N and Levendusky M (2018) The economic consequences of Partisanship in a polar-
ized era. American Journal of Political Science 62, 5–18. ISSN: 15405907.
Mullinix KJ, Leeper TJ, Druckman JN and Freese J (2016) The generalizability of survey experiments. Journal of
Experimental Political Science 2, 109–138. ISSN: 2052-2630.
Palfrey TR and Poole KT (1987) The relationship between information, ideology and voting behavior. American Journal of
Political Science 31, 511–530.
Rogowski JC and Sutherland JL (2016) How ideology fuels affective polarization. Political Behavior 38, 485–508. ISSN:
01909320.
Sniderman PM (2000) Taking sides:a fixed choice theory of political reasoning. In Lupia A, McCubbins MD and Popkin SL
(eds.), Elements of Reason: Cognition, Choice, and the Bounds of Rationality. New York: Cambridge University Press,
pp. 67–84.
Sniderman P and Stiglitz E (2012) The Reputational Premium. Princeton, NJ: Princeton University Press. http://scholar.goo-
gle.nl/scholar?q=Sniderman+and+stiglitz+2012\&btnG=\&hl=en\&as\_sdt=0,5#0.
Tomz M and Van Houweling RP (2009) The electoral implications of candidate ambiguity. American Political Science
Review 103, 83–98.
Webster SW and Abramowitz AI (2017) The ideological foundations of affective polarization in the U.S. electorate.
American Politics Research 45, 621–647. ISSN: 15523373.

Cite this article: Lelkes Y (2019). Policy over party: comparing the effects of candidate ideology and party on affective polar-
ization. Political Science Research and Methods 1–8. https://doi.org/10.1017/psrm.2019.18